Rather than a noisy time series, we get a rather elegant view of an unmistakable trend. It's the same information presented in a more cognitively accessible fashion. The accompanying story is here.
Rather detracting from the value of the image is the fact that the vertical extent of the colored bars is not explained in any detail. It "feels like" some sort of uncertainty measure (older years indeed do have wider bars) but the source of the data and the analysis represented by the bars should be explained somewhere accessible from the page. It would be dramatcially more useful that way.
That said, the image itself is a very nice illustration of Tufte-esque principles, and it makes the actual situation (right, the actual warming trend) much more cognitively accessible than does the usual time series.
Update:Per Hank's suggestion in the comments, here is a tweaked version optimized for the most common form of color blindness:
![](https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEi7Av1j5RJjx9jjl9YPtOG4eKGACeJejR_rJmhqyx6G6lLj1RQ42vifgqWtJ6ZfBmb7i7Kddlf1_Cv4PJYV9TuxmgOOgLyG3VS2i4y5Hbqg6qzePrVioYaASTVGxE967AOKe7kZo6QtjQrY/s400/stripes2.jpg)
3 comments:
Agreeing with the comment at Globalchange
http://groups.google.com/group/globalchange/browse_thread/thread/961f4f77df0d237d?hl=en
Clearly the older bars are taller (and maybe wider?) -- is height the error bar range and width the number of observations, perhaps?
On my screen I'm counting pixels, which gets confusing.
Bad scans, and pixelated screens, can change the apparent proportion of length and height for a little colored rectangle.
Heck, displays can even make a dashed line look like a solid line, to a person looking at a low-resolution copy or a low resolution screen.
Who're you going to trust, the data or your lying eyes?
Is there an RG colourblind-friendly version?
Chuck, this might help:
http://www.vischeck.com/daltonize/
Post a Comment